Mathematical Reasoning on AIME 2025 (AIME25, Math Avg)
48.38AIME25 ScorePSFT
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| PSFT2025.08 | 48.38 | — | |
| SFT2025.08 | 46.88 | — | |
| Granite-3.3 (8B)Variant=PRISM2026.03 | 27.96 | 48.75 | |
| Mistral-Small (24B)Variant=PRISM2026.03 | 27.34 | 47.02 | |
| Mistral-7BVariant=PRISM2026.03 | 24.27 | 41.28 | |
| Granite-4-H Micro (3B)Variant=PRISM2026.03 | 23.49 | 46.64 | |
| Nemotron-H-8k (8B)Variant=PRISM2026.03 | 22.76 | 39.53 | |
| Granite-4 Micro (3B)Variant=PRISM2026.03 | 22.29 | 42.96 | |
| LLaMA-3.1 (8B)Variant=PRISM2026.03 | 19.32 | 36.41 | |
| Granite-4 Micro (3B)Variant=Base2026.03 | 12.34 | 26.28 | |
| Granite-4-H Micro (3B)Variant=Base2026.03 | 2.7 | 13.32 | |
| Nemotron-H-8k (8B)Variant=Base2026.03 | 2.29 | 17.96 | |
| Mistral-Small (24B)Variant=Base2026.03 | 0.73 | 9.48 | |
| Granite-3.3 (8B)Variant=Base2026.03 | 0.31 | 8.95 | |
| LLaMA-3.1 (8B)Variant=Base2026.03 | 0.15 | 2.24 | |
| Mistral-7BVariant=Base2026.03 | 0.1 | 0.59 |